Automated detection of cytochalasin-B blocked binucleated lymphocytes for scoring micronuclei.

A comparison between manual and computer-based automatic scoring of micronuclei (MN) was performed in order to optimize the preparation technique and to validate the image analysis procedure. For this purpose whole human blood of three donors was either irradiated (1 Gy X-rays) or treated with the chemical mutagen methyl methane sulphonate (25 mg/ml) and cultivated in the presence of cytochalasin B to obtain binucleated cells with a high yield of MN. An algorithm for MN detection has been developed for Giemsa (G)- and Feulgen-Congo-Red (FCR)-stained slides. This algorithm contains a sequence of grey operators and binary operators necessary to detect nuclei and MN, and to efficiently reject artefacts. The output is a data file with measurements of cells and intracellular inclusions. From these features, information can be extracted concerning the frequency of the various cell classes (based on nuclearity), the presence of MN and various shape parameters. A close analysis of the automatic scoring of G- and FCR-stained cells, revealed that 59-86% of all automatically classified binucleated cytokinesis-blocked (CB) cells were correctly classified. Although some MN were overlooked during automated scoring, the results show that, on average, similar MN frequencies are obtained with automated and manual scoring. The errors which occurred were mainly due to the misclassification of CB cells, the non-detection of extremely small MN and the aggregation of MN to the main nucleus. The possibility of scanning high numbers of cells overnight, to relocate CB cells with potential MN and the quantitative character of the results offers good prospects for future use in the in vitro MN test.